[NATAP] Pancreatitis in 20 ACTG Studies: incidence/causes

[Guest guest]

NATAP - http;//www.natap.org

Incidence of Pancreatitis in HIV-1-Infected Individuals Enrolled in 20 Adult AIDS Clinical Trials Group Studies: Lessons Learned

JAIDS Journal of Acquired Immune Deficiency Syndromes: Volume 39(2) 1 June 2005

Reisler, B MD, MPH*; , L MD†; Redfield, R MD*; , A ScD‡

From the *Institute of Human Virology, University of land School of Medicine, Baltimore, MD; †Northwestern University, Feinberg School of Medicine, Chicago, IL; and ‡Statistical and Data Analysis Center, Harvard School of Public Health, Boston, MA.

Abstract

Objective: To report on the incidence of clinical- and laboratory-defined pancreatitis in HIV-1-infected individuals treated with antiretrovirals (ARVs).

Methods: Pancreatitis incidence rates were calculated based on a Poisson distribution for subjects enrolled in 1 or more of 20 Adult AIDS Clinical Trials Group studies from October 1989 through July 1999.

Results: A total of 8451 subjects were enrolled. The overall pancreatitis rates were 0.61 per 100 person-years (PYs) clinical and 2.23 per 100 PYs clinical/laboratory. Pancreatitis rates for single, dual, and triple nucleoside reverse transcriptase inhibitors (NRTIs) were similar. Rates of pancreatitis in didanosine (ddI) arms seemed to be dose dependent. Pancreatitis rates in ddI/hydroxyurea (HU) arms were not significantly different from the rates for ddI alone. Overall pancreatitis rates for ddI/stavudine (d4T) trials were high at 4.16 per 100 PYs clinical and 6.25 per 100 PYs clinical/laboratory. The highest rates were seen with the combination of indinavir (IDV)/ddI/d4T with or without HU.

Conclusions: The combination of NRTIs and definition has an impact on the incidence of pancreatitis. Standardization of definition and more comprehensive evaluations are needed to determine how much of this pancreatitis is directly caused by ARVs and how much is attributable to preexisting comorbidities and other known risk factors.

INTRODUCTION

Acute pancreatitis is a potentially life-threatening condition that is characterized clinically by abdominal pain, nausea, and vomiting and biochemically by elevations of lipase and/or amylase. Although the annual incidence in the general non-HIV-infected population is relatively low, estimated to be 17 to 30 cases per 100,000 population,1,2 the annual incidence of acute pancreatitis in the US HIV population is considerably higher. In the pre-highly active antiretroviral therapy (HAART) era, Dutta et al3 reported an acute pancreatitis rate of 14 cases per 100 HIV patients over a one year period. This rate may have been exceedingly high because of the comorbid conditions prevalent in their urban HIV clinic population (eg, ethanol use and biliary disease), the use of medications associated with pancreatitis (eg, pentamidine, high-dose didanosine [ddI], corticosteroids, ketoconazole, sulfonamides, metronidazole, isoniazid), HIV itself, and opportunistic infections (eg, cytomegalovirus, cryptosporidiosis, mycobacterial disease).3,4

In a recent cross-protocol analysis representing a large cohort of nearly 3000 subjects prescribed antiretrovirals (ARVs) in the HAART era, the incidence of grade 4 clinical- and/or laboratory-defined pancreatitis was 0.85 per 100 person-years (PYs).5 Here, we report the incidence of clinical- and clinical/laboratory-defined pancreatitis (grade 3 or 4) by particular ARV treatment regimen in HIV-1-infected individuals enrolled in 20 Adult AIDS Clinical Trials Group (AACTG) clinical trials involving 8451 subjects. This analysis was prompted by the unexpected toxicity findings of severe pancreatitis and death in AACTG 5025, a multicenter, partially blind, prospective, randomized study of 202 HIV-infected subjects, to determine whether therapy intensification with ddI/stavudine (d4T) with or without hydroxyurea (HU) improved long-term rates of virologic suppression. The AACTG terminated 5025 prematurely based on the results of a scheduled interim safety review conducted in September 1999. AACTG 5025 highlights the need to quantify pancreatitis rates observed in other AACTG ARV trials in which ddI was included in at least 1 treatment arm.

RESULTS

Our cross-protocol analysis of 17 studies reflecting 84 study arms (n = 6287) yielded a relatively low overall clinical pancreatitis rate of 0.61 per 100 PYs (n = 58, range: 0.47-0.79) and a higher clinical/laboratory pancreatitis rate of 2.23 per 100 PYs (n = 217, range: 1.95-2.55). Thus, the clinical/laboratory pancreatitis definition yielded a rate nearly 4-fold higher than the clinical pancreatitis definition. These rates reflect a wide range, from 0 per 100 PYs observed in a number of ARV combinations to 14.52 per 100 PYs observed with ddI/d4T/indinavir (IDV)/HU.

Table 2 summarizes the results in the remaining 3 studies (n = 2164) for which pancreatitis incidence rates were calculated using a combined clinical and chemical pancreatitis definition. In these 3 studies, data were managed by the Bristol-Meyers Squibb Company. The range in incidence rates for the various ddI study arms was 1.0 to 14.3 per 100 PYs, whereas the range for the zidovudine (ZDV) study arms was 0.5 to 3.2 per 100 PYs. Rates of pancreatitis in the ddI arms seemed to be dose dependent, with the highest rates in the arms with the highest doses of ddI. It is noteworthy that in these 3 studies, the proportion of subjects with pancreatitis was similar: 9% for AACTG 116, 10% for AACTG 116b/117, and 12% for AACTG 118.

We grouped our pancreatitis rates by ARV combinations in the 17 studies for which data were managed by the SDAC. Of note, the clinical pancreatitis rate for the combination of ddI/ZDV of 0.48 per 100 PYs (n = 11, range: 0.24-0.86) was similar to that of ddI monotherapy of 0.38 per 100 PYs (n = 5, range: 0.12-0.88) and ZDV monotherapy of 0.28 per 100 PYs (n = 4, range: 0.08-0.72). In addition, when using the clinical/laboratory definition, the combination ddI/ZDV rates were not significantly different from the rates for ddI monotherapy. When using the same clinical/laboratory pancreatitis definition for ddI/ZDV, however, a significantly higher rate of 2.34 per 100 PYs (n = 54, range: 1.76-3.19) was observed as compared with ZDV monotherapy of 0.91 per 100 PYs (n = 13, range: 0.49-1.56) (P < 0.01). The ddI/ZDV pancreatitis rates for clinical and clinical/laboratory were similar to the dideoxycytidine (ddC)/ZDV rates for clinical of 0.69 per 100 PYs (n = 11, range: 0.34-1.23) and clinical/laboratory of 1.81 per 100 PYs (n = 29, range: 1.21-2.60).

The ddI/HU combination demonstrated a low rate of clinical pancreatitis of 0 per 100 PYs (n = 0, range: 0-3.95) and a higher rate for clinical/laboratory pancreatitis of 3.21 per 100 PYs (n = 3, range: 0.66-9.39). These rates were not significantly different from the corresponding rate for ddI alone. ZDV alternating with ddI yielded a slightly higher than expected rate for clinical pancreatitis of 1.66 per 100 PYs (n = 5, range: 0.54-3.88) and clinical/laboratory pancreatitis of 3.99 per 100 PYs (n = 12, range: 2.06-6.96). For clinical pancreatitis, this rate was not significantly different from that of ddI monotherapy, ZDV monotherapy, or combined ddI/ZDV dual nucleoside reverse transcriptase inhibitor (NRTI) therapy. The clinical/laboratory pancreatitis rates for ZDV alternating with ddI were significantly higher than for ZDV monotherapy alone, however (P < 0.01). The combination of nevirapine (NVP)/ZDV/ddI (see Table 3C) demonstrated a significantly higher rate of clinical pancreatitis and clinical/laboratory pancreatitis as compared with ZDV/ddI (clinical, P < 0.05; clinical/laboratory, P < 0.001). Compared with delavirdine (DLV)/ZDV/ddI, the rates for NVP/ZDV/ddI were not significantly different for clinical pancreatitis or clinical/laboratory pancreatitis.

Although our analysis contained relatively few protocols with protease inhibitor (PI)-based regimens, it seems that PI-based regimens of NVP or IDV combined with 3TC/d4T, ZDV/3TC, ddI/3TC, or ddI/ZDV yielded similar pancreatitis rates to dual nucleoside regimens: 0.42 per 100 PYs (n = 1, range: 0.01-2.32) for clinical pancreatitis and 1.10 per 100 PYs (n = 3, range: 0.23-3.21) for clinical/laboratory pancreatitis. When NVP or IDV was combined with ddI/d4T, however, high rates of pancreatitis were observed, with IDV/ddI/d4T representing the combination with the highest rates (see Table 3D).

Table 4 provides the pancreatitis incidence rates for all treatment regimens that included ddI/d4T. The overall clinical rate was 4.16 per 100 PYs (n = 9, range: 1.90-7.90), and the overall clinical/laboratory rate was 6.25 per 100 PYs (n = 20, range: 3.82-9.65). The highest rates were seen with the combination of IDV/ddI/d4T. The addition of HU to IDV/ddI/d4T did not have a significant impact on the incidence rate of pancreatitis.

The use of recombinant human interleukin-2 (rhIL-2) did not seem to increase the risk of pancreatitis. Among all study arms that included rhIL-2 as part of the treatment regimen, reflecting 128 PYs of follow-up, the rate of clinical pancreatitis was 0 per 100 PYs (n = 0, range: 0-2.88) and the rate of clinical/laboratory pancreatitis was 0 per 100 PYs (n = 0, range: 0-2.74). Among the 3 studies that failed to exclude subjects with a history of pancreatitis (AACTG 298, 328, and 5025), high pancreatitis rates were only observed in AACTG 5025.

DISCUSSION

This analysis represents the largest effort to date to link the incidence of pancreatitis to particular antiretroviral regimens systematically. Our principal findings are that the combination of NRTIs selected and the pancreatitis definition chosen seem to have an impact on the incidence of pancreatitis. Of the various combinations of dual NRTIs that we studied, ddI/d4T seems to be associated with the highest rates of pancreatitis. The combination of IDV/ddI/d4T seems to be associated with particularly high rates of pancreatitis, reminiscent of high-dose ddI monotherapy trials. Nonetheless, we did observe a great deal of variability in pancreatitis rates among the various study arms.

The clinical pancreatitis definition probably underestimates the incidence rate (emphasizing specificity), and the laboratory or chemical pancreatitis definition probably overestimates the incidence rate (emphasizing sensitivity), especially in studies using amylase levels as the chemical measure of pancreatitis. High levels of serum amylase may reflect high levels of S-type isoamylase secreted chiefly by salivary glands, as seen commonly with salivary hyperplasia and parotid disease.6 Hyperamylasemia has also been associated with renal insufficiency, liver disease, gastrointestinal malabsorption, female reproductive tract disease, acidosis, macroamylase, cancer, and HIV.6-8 Similarly, hyperlipasemia has been associated with renal insufficiency,9 liver disease,8,10 gastrointestinal malabsorption,11 acidosis,12 macrolipase,13 cancer,14 and HIV.8

Although the serum lipase assay typically has greater sensitivity and specificity for pancreatitis than the serum amylase assay, discrepancies among the different methodologies have been reported.15 These discrepancies may be attributable to the patient, reflecting moderately high biologic variation as a result of the particular laboratory assay and methods for calibration, or to an interaction of the patient's sera with the particular lipase assay. Calleja and Barkin16 have suggested that to improve the specificity of the lipase assay, one should require an increase of 3 times the ULN. This suggestion is supported by the observations of Byrne et al17 that mild elevations of pancreatic enzymes (less than 3 times the ULN) were not associated with significant pathologic findings. We used the NIH National Institute of Allergy and Infectious Diseases (NIAID) Division of AIDS (DAIDS) grade 3 and 4 amylase and lipase levels for our analysis, which reflect a range of 2 to 5 times the ULN and are in line with these suggestions.

The frequency of ddI-induced pancreatitis seems to be dose related.18,19 This observation is supported by the pancreatitis rates in the 3 early AACTG studies (AACTG 116a, 116b/117, and 118) in which the pancreatitis definition was the same based on combined clinical and laboratory values. In these studies, data were collected and analyzed in a similar fashion and CD4 cell count entry criteria were similar. Subjects in AACTG 116b/117 had significantly lower rates of pancreatitis than subjects in AACTG 116a and 118. Perhaps this variability may be explained by differences not readily apparent in the patient population or in the way data were collected. Another possible explanation is based on the differences in length of follow-up in these 3 studies. In these 3 studies, the proportion of subjects with pancreatitis was similar, suggesting that pancreatitis may have been an early event. Thus, the study with the longest follow-up, (AACTG 116b/117) had lower incidence rates than the studies with shorter follow-up rates (AACTG 116 and 118).

The observation that high ddI plasma levels may be associated with higher rates of pancreatitis was recently seen once again when clinicians began using 400 mg of ddI daily in combination with 300 mg of tenofovir daily.20 It is now well established that coadministration of tenofovir with ddI increases the maximum plasma concentration and area under the curve (AUC) of ddI by 48% to 64%.21 One ARV combination that demonstrated surprisingly high rates of pancreatitis was NVP/ZDV/ddI. This regimen should probably be used with caution in patients at increased risk of pancreatitis.

The package inserts for ddI and d4T contain black box warnings with regard to the potential risk of pancreatitis.22,23 In fact, the d4T package insert black box warning issues a specific warning with regard to the combination of ddI plus d4T, with or without HU.23 In regard to the combination of ddI and d4T, we have 7 treatment arms in 3 studies of interest: AACTG 328, 364, and 5025. Our findings support the black box warning with regard to this combination. Clearly, of all the nucleoside combinations included in our analysis, the combination of ddI/d4T was associated with the highest rates of pancreatitis. The need for adequate sample size and follow-up to detect significant clinical toxicities is underscored by the observation that there were individual study arms containing the combination of ddI/d4T (eg, AACTG 328) that yielded no pancreatitis, most likely as a result of the limited sample size and follow-up.

In their AACTG 5025 pancreatitis case-control analysis, Havlir et al24 were unable to identify factors that predisposed the AACTG 5025 subjects to pancreatitis other than ARV therapy with IDV/ddI/d4T with or without HU.24 Havlir et al24 do point out that their case-control analysis was limited by sample size and suboptimal baseline lipid parameters, namely, nonfasting cholesterol and triglycerides. Yet, it is noteworthy that of the 3 subjects who died, 2 had high baseline triglycerides (1205 mg/dL and 770 mg/dL) and the third had a history of cerebrovascular disease. Among these 3 subjects, baseline comorbidity seemed to be a key contributing factor.

One distinct possibility is that the regimen of ddI/d4T/IDV induces particular metabolic and lipid abnormalities at the cellular level that significantly increase the risk of pancreatitis. Another possibility is that the regimen might lead to biliary tract sludge formation and/or biliary microlithiasis and, consequently, biliary pancreatitis, as has been observed with ceftriaxone.25 It is well known that ddI/d4T has been associated with hepatic steatosis23 and that IDV is associated with an elevated unconjugated hyperbilirubinemia.26 It stands to reason that if ddI/d4T causes hepatic dysfunction, it may lead to impairment of IDV metabolism and, consequently, an increase in biliary IDV levels. High IDV levels, coupled with a slight increase in biliary pH because of increased levels of unconjugated bilirubin in bile, might predispose to the precipitation of IDV crystals and/or calcium bilirubinate crystals in the biliary tree.

Our analysis includes only 2 clinical trials that used HU: AACTG 307 and 5025. These studies provide us with some important insights into the use of HU in combination with ddI and ddI/d4T. In AACTG 307, the pancreatitis rates for those taking ddI/HU were not different from those taking ddI alone. Similarly, the pancreatitis rates in AACTG 5025 for those taking IDV/ddI/d4T/HU were not different from those taking IDV/ddI/d4T. Although all 3 deaths occurred in the IDV/ddI/d4T/HU arm, the study inclusion and exclusion criteria did not adequately address known risk factors for pancreatitis (eg, baseline triglycerides) and prior history of pancreatitis.24

The absence of a consensus pancreatitis definition, inconsistencies in the manner in which clinical and laboratory data are collected and analyzed, and the paucity of long-term follow-up have resulted in widely variable pancreatitis incidence rates. This, in turn, has resulted in inconsistencies with regard to predictors of pancreatitis in HIV patients. In their retrospective analysis, et al27 suggest that significant risk factors for pancreatitis include use of HU, CD4 count <200 cells/mm3, female gender, and history of pancreatitis. Ethanol use, biliary tract disease, concomitant medications, and hypertriglyceridemia, all well-known risk factors for pancreatitis, were not considered in their analysis.

We were unable to perform a cross-protocol analysis of incidence and risk factors, because potential risk factor data were not collected in a uniform and systematic fashion. The inherent nature of US clinical trials study teams (the AACTG included) is to address the efficacy and safety of particular drug regimens. Although this lack of uniformity perhaps provides an effective mechanism by which to assess the efficacy of specific drug combinations, it contributes to a lack of generalizability with regard to safety data. In this vein, Ioannidis and Lau28 have published extensively on the overall inadequacy of safety reporting in published clinical trials. In addition, our ability to compare different pancreatitis rates by ARV treatment combinations was limited to some extent by sample size, duration of follow-up, and need to adjust for multiple comparisons.

Our study has several limitations. With regard to the clinical pancreatitis definition, the diagnosis of pancreatitis was not based on a diagnostic procedure; an abdominal CT scan, a surgical procedure, or histopathologic evaluation would be more definitive. These diagnostic procedures are rarely performed in cases of mild pancreatitis, however. Moreover, the studies did not systematically collect data with regard to diagnostic procedures. Consequently, in a review of previous clinical trials, we were unable to control for how clinical pancreatitis was defined and diagnosed. Our clinical pancreatitis definition is predicated on the study investigators' reporting of the diagnosis. This deficiency resulted in the formalization of a diagnosis of pancreatitis with the AACTG so that this problem can be addressed with more clarity in the future.29 Clinical pancreatitis events were not reviewed by a central review committee of medical experts. Patients were not screened with a diagnostic test (eg, right upper quadrant ultrasound examination, CT of the abdomen, endoscopy, oral cholecystography) before protocol enrollment. Finally, there is a paucity of pancreatitis incidence data among HIV-uninfected individuals and HIV-infected individuals who are antiretroviral treatment naive with similar demographics for comparison.

Better evidence is needed to determine how much of this pancreatic morbidity is directly attributable to ARV therapy and how much is related to preexisting comorbidity. To assess the impact of ARVs as well as other agents (eg, rhIL-2, HU) on the incidence of pancreatitis, we need to capture more comprehensive data with regard to comorbid conditions (eg, hypertriglyceridemia, liver disease, gallbladder disease, renal disease), previous history of pancreatitis, body mass index (BMI), recreational drugs and alcohol, concomitant medications at baseline and throughout follow-up, smoking, and genetic predispositions. This, coupled with a pancreatitis case definition that is current and easily implemented, should facilitate cross-protocol analysis and risk factor analysis. This approach would help to ensure that we arrive at appropriate conclusions with regard to ARV-related pancreatitis.

METHODS

The AACTG is a large National Institutes of Health (NIH)-funded clinical trials group that conducts research through a national network of academic-based clinical sites. Twenty AACTG studies are included in this analysis: AACTG 116a, 116b/117, 118, 143, 175, 193a, 194, 241, 244, 261, 276, 290, 298, 302, 303, 306, 307, 328, 364, and 5025. All 20 studies had at least 1 treatment arm that contained ddI. Active enrollment spanned a 10-year period from October 1989 through July 1999. AACTG 175, 302, 303, and 364 include the same subjects. AACTG 302 and 303 were rollover studies for AACTG 175, and AACTG 364 was a rollover study for AACTG 302 and 303. Data for 3 studies (AACTG 116a, 116b/117, and 118) were managed by a contractor for the pharmaceutic company (Bristol-Meyers Squibb Company), and data for the remaining 17 studies were managed by the Statistical and Data Analysis Center (SDAC), Harvard University, School of Public Health, Boston, MA.

In AACTG 116a, 116b/117, and 118, pancreatitis was defined as elevated serum amylase and a compatible clinical syndrome of nausea, vomiting, and/or abdominal pain. Elevated amylase was to be confirmed as pancreatic in origin with an elevated fractionated amylase or lipase determination. Thus, an individual with minimally elevated amylase (grade 2 or higher) with a clinical presentation consistent with pancreatitis would be classified as a case of clinical pancreatitis. For these studies, only 2 pancreatitis-related variables (time to pancreatitis and an indicator of developing pancreatitis) were available in the database. For the remaining 17 studies managed by the SDAC, 2 definitions of pancreatitis were considered: clinical pancreatitis and a combined definition of clinical and/or laboratory (chemical) pancreatitis. Chemical pancreatitis was defined by significant amylase and/or lipase abnormalities. Clinical pancreatitis cases were found by searching the data from the diagnoses form, death form, and adverse event related (AER) form. This procedure was uniform for all studies. There were no uniform criteria used to make the diagnosis of clinical pancreatitis, however, and there was no committee that reviewed these events. Clinical pancreatitis ranged in severity from grade not reported to grade 5. The clinical pancreatitis definition is predicated on the study investigators' accurate reporting of the diagnosis. Chemical or laboratory pancreatitis, defined as grade 3 or 4 amylase and/or lipase elevation, was found by searching the chemistry and hematology results form(s). The frequency at which amylase and/or lipase was checked was protocol specific, but intervals between testing ranged from a 2- to 16-week period. Most protocols tended to check safety laboratories more frequently during the first 16 weeks of study participation. Patients were not screened with a right upper quadrant ultrasound examination, computed tomography (CT) of the abdomen, endoscopy, or oral cholecystography before protocol enrollment.

In this analysis, study treatment refers to a subject's randomized treatment and may not reflect the actual treatment the subject was receiving. Subjects who never started study treatment were excluded from analyses. With regard to AACTG 116a, 116b/117, and 118, the occurrence of pancreatitis was censored 30 days after the date of the last dose of treatment or the date of treatment crossover, whichever came first. For the remaining 17 studies, with regard to clinical pancreatitis, follow-up began with treatment initiation and ended with the earlier of the last dose of study treatment plus 28 days (4 weeks), date of treatment crossover, or last contact date. With regard to laboratory pancreatitis, data were censored at the earlier of the last dose of study treatment plus 28 days (4 weeks), date of treatment crossover, or last date at which a laboratory value was collected.

Seventeen of 20 studies excluded subjects with a history of pancreatitis on enrollment. The studies that failed to exclude subjects with a history of pancreatitis were AACTG 298, 328, and 5025. Eighteen of the 20 studies excluded subjects for elevation of amylase, pancreatic amylase, and/or lipase at entry. AACTG 116a, 116b/117, and 118 required that subjects' amylase at entry be ≤1.3 times the upper limit of normal (ULN). AACTG 143 required that subjects' amylase at entry be within the ULN. AACTG 175, 241, 244, 261, 276, 290, 302, 303, 306, 364, and 5025 required that subjects' amylase be ≤1.5 times the ULN (grade 1 or less) or that lipase or pancreatic amylase be ≤1.5 times the ULN (grade 1 or less). AACTG 193A, 194, and 307 required that subjects' amylase be ≤2 times the ULN (grade 2 or less) or that lipase or pancreatic amylase be ≤2 times the ULN (grade 2 or less). Patients were not excluded if they had a history of biliary tract disease or heavy alcohol ingestion.

Only the first pancreatitis incident is analyzed. Incidence rates (per 100 PYs) and 95% confidence intervals are calculated based on a Poisson distribution (ie, the incidence rate in any interval of time depends only on the length of the time interval). P values for comparisons of different pancreatitis rates were adjusted for multiple comparisons using a Bonferroni adjustment. This adjustment was made for 4 sets of structured comparisons and allowed for the comparison of the clinical events and the combined clinical/laboratory events.